Add render command for precomputed dashboard data artifacts

Problem

Add a CLI path for rendering from a precomputed JSON artifact so the final render stage can be rerun without recompiling or requerying. Define the artifact contract, decide whether this belongs as a dft render input option or a separate command, and make it useful for eval workflows, cached runs, and reproducible debugging of rendering output.

Context

Use cases

This enables three things: 1. Eval workflows — render eval analysis dashboards over JSONL results without needing a live database connection. Related to mcp-analyst-agent/persist-eval-outputs-for-dataface-analysis-and-boards. 2. Cached/offline rendering — re-render a dashboard from saved query results, useful for CI artifacts, sharing, and debugging. 3. Reproducible render debugging — isolate render-layer bugs by replaying the same data through the renderer without recompiling or requerying.

Current pipeline

dft render currently runs the full pipeline: YAML → compile → execute queries → render. This task adds an entry point that skips compile+execute and starts from a precomputed data artifact.

Priority note

This is P2 but becomes more valuable if the eval dashboard use case (task 5 in mcp-analyst-agent) needs to work without a live query target. For M1, if eval dashboards can point at local JSONL files via DuckDB, this task may not be blocking. Reassess priority after the eval runner output format is finalized.

Possible Solutions

Plan

Implementation Progress

QA Exploration

• QA exploration completed (or N/A for non-UI tasks)

N/A for browser QA. Validation: render a dashboard from a saved artifact, compare output to the same dashboard rendered through the full pipeline.

Review Feedback

• Review cleared